PHILOSOPHY WITHOUT BELIEF

Zach Barnett
Abstract: Should we believe our controversial philosophical views? Recently,
several authors have argued from broadly conciliationist premises that we
should not. If they are right, we philosophers face a dilemma: If we believe
our views, we are irrational. If we do not, we are not sincere in holding them.
This paper offers a way out, proposing an attitude we can rationally take to-
ward our views that can support sincerity of the appropriate sort. We should
arrive at our views via a certain sort of ‘insulated’ reasoning – that is, reason-
ing that involves setting aside certain higher-order worries, such as those
provided by disagreement – when we investigate philosophical questions.

Here is what seems to be a fact about our discipline: Some of us really believe
the controversial philosophical views we advocate.1 Some of us really believe
that it can sometimes be rational to have inconsistent beliefs, that seemingly
vague predicates must have precise application conditions, or that a person
would survive if each of her brain cells were replaced with an artificial func-
tional duplicate.
Here is another fact about our discipline: There is widespread disagree-
ment among philosophers surrounding these issues.2 Given certain assump-
tions about the nature of these philosophical disagreements, and given cer-
tain assumptions about the epistemic import of disagreement more generally,
one might come to doubt that our controversial philosophical beliefs are ra-
tional – insofar as we have them. Indeed, numerous authors have developed
arguments along these lines.3 The details of their arguments need not concern
us, but it will be useful to examine briefly one argument in outline, which
will serve as a representative simplification of what they have said:
Conciliationism: A person is rationally required to withhold belief in the face of
disagreement – given that certain conditions are met.4
Applicability: Many disagreements in philosophy meet these conditions.

No Rational Belief: Philosophers aren’t rational to believe many of their contro-

versial views.

1 DeRose (forthcoming) argues that we do not genuinely believe our controversial views

in philosophy, offering an intriguing story about what we might be doing instead. While I
suspect that at least some of us do genuinely believe our controversial views, the arguments
given here do not depend on any such assumption.
2 The Bourget and Chalmers (2013) survey asks philosophers for their views on thirty

questions that are taken to be central to the field. For virtually all of these, we do not observe
anything like consensus.
3 See Brennan (2010), Christensen (2014), Fumerton (2010), Goldberg (2009, 2013a), Korn-

blith (2010, 2013), Licon (2012).

4 Although there is some debate over how these conditions should be characterized, it suf-

fices for us to note that they typically involve the person’s having good reason to consider her
disagreer(s) equally trustworthy, with respect to the disputed sort of issue, as she considers
herself (and her agreers).

-1-
The first premise, Conciliationism, enjoys ample precedent.5 Its strengths and
weaknesses have been thoroughly explored. I will not rehearse the debate
here. The second premise, Applicability, is somewhat less familiar, however,
so it may be helpful to see what has been said about it. Here is how Christen-
sen motivates the position:
I do have good reason to have as much epistemic respect for my philosophical
opponents as I have for my philosophical allies and for myself… In some cases, I
have specific information about particular people, either on the basis of general
knowledge or from reading or talking to the particular epistemologists in ques-
tion. [...]
But another reason derives from the group nature of philosophical contro-
versy. It seems clear that the groups of people who disagree with me on various
philosophical issues are quite differently composed. Many who are on my side of
one issue will be on the other side of different issues. With this structural feature
of group disagreement in philosophy in mind, it seems clear that it could hardly
be rational for me to think that I’m part of some special subgroup of unusually
smart, diligent, or honest members of the profession. (Christensen 2014, p. 146)

Kornblith takes a similar view, at least with respect to one specific debate:
Disagreements within philosophy constitute a particularly interesting case...
Consider the debate between internalists and externalists about epistemic justifi-
cation. I am a committed externalist. I have argued for this position at length
and on numerous occasions. [...]
At the same time, I recognize, of course, that there are many philosophers
who are equally committed internalists about justification[.] It would be reassur-
ing to believe that I have better evidence on this question than those who disa-
gree with me, that I have thought about this issue longer than internalists, or that
I am simply smarter than they are, my judgment superior to theirs. It would be
reassuring to believe these things, but I don’t believe them; they are all manifest-
ly untrue. (Kornblith 2010, p. 31)

In light of these observations, Applicability, too, can seem to be a fairly at-

tractive position.6
This paper takes both Conciliationism and Applicability for granted
(along with their consequence, No Rational Belief) in order to investigate
what sense we can make of philosophy if they are true. If we philosophers –
in an effort to be more rational – suddenly decided to withhold belief about
all philosophically controversial matters, would the practice of philosophy be

5 See Feldman and Warfield (2009), or Christensen and Lackey (2013) for helpful discus-
sions of this principle.
6 An opponent of Applicability is Grundmann (2013). Grundmann points out that even if

my opponent is, in general, as philosophically competent as I am (equally honest, equally in-

telligent, equally hardworking, etc.), she may not be as reliable as I am with respect to the dis-
puted issue. The observation is a good one. But it is worth noting that these general competen-
cies can still serve as fallible indicators of one’s domain-specific reliability. Grundmann does
not argue that there is no correlation between these distinct competencies.

-2-
in any way diminished? As we will see, there is some cause for concern.7

1 The Sincere Philosopher’s Dilemma

On the face of it, it is not immediately clear why giving up our philosophical
beliefs should lead to any problems. Take competitive debate. It is surely
quite common for a debater to defend a view she does not, strictly speaking,
believe. And this fact hardly serves to undermine the practice of debate. Why
should it be any different in philosophy? Perhaps we arrive at certain views,
somehow or other, and then defend them as ably as we can manage – with-
out necessarily believing them to be true. Might this be a reasonable way for
philosophy to operate?
I find myself somewhat uncomfortable with this picture. In particular, it
seems to me that if philosophy were to operate this way, something im-
portant would be missing: namely, the sincerity with which we defend our
preferred positions – a distinctive kind of sincerity that is often lacking in the
context of competitive debate.
The kind of sincerity that I have in mind is a way in which I suspect many
philosophers identify with the views they defend. The thought is that, for
many of us, our views seem right to us, in some important sense. When I re-
flect on the relevant issues, my thinking leads me to certain conclusions. And
if I were, for some reason, obligated to defend other views (perhaps because
they were assigned to me by some governing body), these other views would
not be as sincerely held. In defending the assigned views, I would not neces-
sarily be calling the shots as I see them; my own thinking would not have led
to them.
This is supposed to capture, intuitively, what it takes for one’s views to be
sincerely held.8 My claim is not that philosophers ought to hold their views sin-
cerely, but rather, that many philosophers do experience this sincere com-
mitment toward their favored views and would prefer to be able to continue
doing philosophy in this sincere manner. And this may be more than just a
personal preference for the feeling of sincerity. It is plausible that for many of
us, doing philosophy well – energetically and creatively – comes most natu-
rally when we do sincerely identify with the views we defend.
If this is right, then we ‘sincere’ philosophers have a potentially serious
problem on our hands. There seems to be a tension between this sincerity de-
sideratum, on the one hand, and No Rational Belief, on the other. We can put
the point as a dilemma:
Sincere Philosopher’s Dilemma: Either we philosophers will believe our

7 The following discussion is indebted to Goldberg (2013b). Goldberg’s view will be dis-
cussed in detail later.
8 My choice of the word ‘sincere’ should not be taken to indicate that philosophers who

lack this feeling toward their views are being insincere, in some problematic way. I am simply
pointing out a way in which many of us identify with the views we defend.

-3-
 controversial views or we will not. If we do, then we will be irrational. If
we do not, then our views will not be sincerely held.
The main task of this paper is to show how this challenge can be met. But
first, the challenge should be strengthened. The worry that gives rise to the
challenge is that belief is required for the relevant kind of sincerity. But this
claim is probably too strong.
To see this, assume a Lockean account of belief, according to which out-
right belief just is confidence above a certain threshold, say .75. Let us imag-
ine that I often spend my time working out difficult math problems, replete
with tempting pitfalls that frequently trip me up. Over the years, my success
rate on these problems is only .74, and I know this fact about my reliability.
As a result, when I arrive at an answer to any one of these problems, my con-
fidence in the answer I reach tends not to be quite high enough for belief. De-
spite my lack of outright belief in my answer, the answer I arrive at still
seems right to me, in an important sense. My own thinking led me to it. And
even though I recognize that there is a good chance I erred, overall, I regard
my answer as more likely to be correct than not. In such a case, I find it natu-
ral to say that my commitment to the answer I arrived at is sincere in the rel-
evant sense. If this is right, then outright belief in one’s view is not necessary
for sincerity.
Perhaps this is right. Even if so, it seems to me that the dilemma propo-
nent need not be terribly concerned, for she can reply as follows:
Perhaps I was too quick in suggesting that outright belief was the only doxastic
attitude capable of supporting sincerely held views. A fairly high credence prob-
ably can do the trick. But this observation hardly saves the sincere philosopher,
for it is doubtful that we can rationally maintain high credences in our controver-
sial views. The same considerations that gave rise to No Rational Belief (i.e. Con-
ciliationism, Applicability) are sure to entail a parallel No Rational High Confi-
dence principle, which will forbid the high credences present in the alleged
counterexample. So here’s a more general challenge: Tell us specifically which at-
titude you will take toward controversial views that can get you both rationality
and sincerity.

This reply seems to me to be exactly right. The challenge is not simply to

demonstrate how to achieve sincerity without belief, but rather, to demon-
strate that there is some attitude, or some set of attitudes, which allow for
sincere and sensible participation in philosophy. The next section examines
one potential answer, due to Sanford Goldberg.

2 Philosophical Views as Speculations

Goldberg has explored nearby territory in a series of recent papers (2013a,
2013b). He defends a version of No Rational Belief, and so he is concerned
with a question similar to the one we are considering. He writes:
Unless we want to condemn philosophers to widespread unreasonableness (!),

-4-
 we must allow that their doxastic attitude towards contested propositions is, or
at any rate can be, something other than that of belief. (Goldberg 2013b, p. 282)

Though Goldberg is not explicitly concerned with allowing that philosophers

can sincerely hold their views in the sense discussed in the previous section,
he is sensitive to nearby issues, such as the sincerity of philosophical asser-
tion.
Goldberg thinks that there is indeed an attitude that we philosophers can
and should rationally take toward our views: ‘[T]here is an attitudinal cousin
of belief which is reasonable to have even under conditions of systematic dis-
agreement and which captures much, if perhaps not all, of the things that are
involved in “having a view” in philosophy’ (Goldberg 2013b, p. 284). The rel-
evant state is called ‘attitudinal speculation’:
Speculation: [O]ne who attitudinally speculates that p regards p as more likely
than not-p, though also regards the total evidence as stopping short of warranting
belief in p. (Goldberg 2013b, p. 283)

Goldberg goes on to suggest that this attitude is what is required for sincere
and proper assertion in the context of philosophy. The picture of philosophy
being recommended (henceforth, ‘the speculation picture’) is, I think, a fairly
natural one: Advocates of Incompatibilism, say, should hold their view to be
more likely than its rival; at the same time, they should acknowledge that the
total evidence (including evidence from disagreement) does not permit suffi-
cient confidence in Incompatibilism for outright belief.
This picture is most attractive when applied to philosophical issues that
divide philosophers into exactly two camps. Goldberg’s picture may require
refinement, however, in order to handle debates consisting of three or more
rival positions. For example, take normative ethics. Oversimplifying dramati-
cally, let us suppose that Consequentialism, Deontology, and Virtue Ethics
are equally popular, mutually exclusive views exhausting the plausible op-
tions. According to the speculation picture, my being a Deontologist will re-
quire me to have a credence in Deontology exceeding .5. There are two poten-
tial worries here.
First, it is not clear that such a high level of confidence in Deontology can
be rationally maintained, on a broadly Conciliationist picture. Admittedly,
the version of Conciliationism discussed in the introduction, which dealt
with all-or-nothing beliefs, would be silent about this question. But some ver-
sions of Conciliationism do place rational constraints on one’s level of confi-
dence, and in general, they would demand that one’s level of confidence in
Deontology be roughly ⅓ in a case like this.9

9 In the case described, the three views were supposed to be about equally popular. So
about ⅔ of my peers reject Deontology and opt for one of the other two views. If I have every
reason to think these opponents are, in general, as philosophically reliable as my fellow Deon-
tologists (and myself), then I will not be rationally permitted (from within a broadly Concilia-

-5-
 Second, one of the most attractive features of the speculation picture — its
ability to allow us to favor our own view very slightly — seems to disappear
in cases like this one. Suppose that my confidence in Deontology is .4, while
my confidence in each alternative is .3. It would seem to make sense to classi-
fy me as a Deontologist, but, according to Goldberg’s view, this would be a
mistake. My confidence in Deontology is, apparently, too low. So the attrac-
tive feature of the speculation picture disappears in cases like this.
With the foregoing problems in mind, we might modify Goldberg’s view
as follows:
Speculation*: One who attitudinally speculates that p regards p as the likeliest
option (given some set of options), though also regards the total evidence as
stopping short of warranting belief in p.

This amended version seems to capture the spirit of Goldberg’s proposal

nicely, allowing us to lean slightly toward our preferred positions even when
there are multiple incompatible ones on offer.10 Can the revised account pro-
vide an answer to our dilemma? Specifically, can speculation* be the doxastic
attitude underlying our philosophical commitments?

3 Obstacles for the Speculation Account

In assessing his own account, Goldberg points out that attitude of speculation
may not be sufficient for having a view in philosophy, since proponents of a
philosophical view are ‘typically more motivated to persist in defense of the
view when challenged, than is one who merely speculates that p’ (Goldberg
2013b, p. 284). In response, he suggests that speculation should be taken to be
a necessary condition on having a philosophical view, not a sufficient one
(Goldberg 2013b, p. 284).
But there is reason to worry that speculation (and even speculation*),
may not be a necessary condition, either. There seem to be cases in which one
can sensibly have a view, despite regarding it as less likely, all things consid-
ered, than some rival view. Anticipating this complaint, Goldberg asks
whether it ever makes sense for one to defend a view she regards as a ‘long
shot.’ Ultimately, he suggests, though, that there is something ‘slightly per-
verse’ about one’s holding a view even when she does not think that the view

tionist framework) to think my side is more likely right than not on this occasion. See, e.g.,
Elga (2007).
10 It is also worth noting that speculation* incurs a problem of individuation from which

the original speculation is immune. Suppose that my confidence in Consequentialism is .4 and

that my confidence in each of the others is .3. Since I regard Consequentialism as likelier than
the other options, it seems clear that I do take the attitude of speculation* toward Consequen-
tialism. But we might carve the options up differently: If instead we say that there are two
views on the table – Consequentialism and non-Consequentialism – then I cannot be said to
take the attitude of speculation* toward Consequentialism after all. We can set this difficulty
aside, however, for I will suggest that both versions of the speculation picture are susceptible
to a more pressing problem.

-6-
will, in the end, be better-supported by the total evidence (Goldberg 2013b, p.
283 fn. 5). While I share the intuition, thinking about certain cases suggests to
me that this is not as problematic as it might seem. Consider an analogy:
Logic Team: You are on a five-player logic team. The team is to be given a logic
problem with possible answers p and not-p. There is one minute allotted for each
player to work out the problem alone followed by a ten-second voting phase,
during which the team members vote one by one. The answer favored by a majori-
ty of your team is submitted.
You arrive at p. During the voting phase, Vi, who is generally more reliable
than you are on problems like these, votes first, for not-p. You are next. Which
way should you vote?

Given a broadly Conciliationist view, it is not rational for you to regard your
answer of p as more likely than its negation, after seeing Vi vote. But there is,
I think, still pressure on you to vote for the answer you arrived at, rather than
the one you now regard as most likely to be correct.
We can illustrate this by adding a bit more information to the case. Sup-
pose that Vi’s reliability is .9, that the reliability of each other team member is
.75, that each team member is statistically independent of each other, and that
the team is aware of this information. If everyone were to defer to Vi during
voting, the team would perform sub-optimally in the long run.11 So in this
collaborative truth-seeking context, there is nothing troublesome about ‘de-
fending’ a view while thinking that it is more likely incorrect than not. More
generally, we can see that, in this context at least, one’s all-things-considered
confidence is no sure guide to what view one should put forward as one’s
own.
Does this point carry over to philosophy? Perhaps. Within a broadly Con-
ciliationist framework, how popular a position is (among some group of
trustworthy evaluators) partly determines how much confidence one should
have in that position, all things considered. But it seems doubtful that the
philosophical popularity of a view should have much impact on whether a
given philosopher should hold that view herself. Consider an example.
Turning Tide: Physicalism seems right to Pat. She finds the arguments for Physi-
calism to be persuasive; she is unmoved by the objections. And at present, Physi-
calism is the most popular view among philosophers of mind/metaphysics. On
balance, she considers herself a Physicalist.
Later, the philosophical tide turns in favor of Dualism. Perhaps new argu-
ments are devised; perhaps the familiar objections to Physicalism simply gain
traction. In any case, Pat remains unimpressed. She does not find the new argu-
ments for Dualism to be particularly strong, and the old objections continue to
seem as defective to her as they always have.

11Following this strategy, the team’s reliability would just be Vi’s reliability: .90. If each
team member votes without deferring, the team’s reliability can be shown to be considerably
higher: approximately 0.93.

-7-
What should Pat’s view be, in a case like this? If Pat is a Conciliationist who
happens to regard other philosophers of mind and metaphysics as generally
trustworthy about philosophical matters (which we’ll suppose she is), her all-
things-considered confidence in Physicalism may well decrease as Dualism
becomes the dominant view, perhaps dipping below .5. But what seems
strange is that once Pat’s all-things-considered confidence in Physicalism falls
low enough, and once her all-things-considered confidence in Dualism rises
high enough, she should stop being a Physicalist (and perhaps become a Du-
alist) — solely on the basis of its popularity, and despite that, when she thinks
about the relevant arguments and objections, Physicalism still seems more plausible
to her. Perhaps there is a special role for one’s own consideration of the issues
to play in contexts like these.

4 Disagreement-insulated Inclination
Thinking about the preceding examples suggests a different approach alto-
gether: As a philosopher, my views should be informed only by the way that
some of the evidence seems to me to point. In particular, I should set aside the
evidence I get from the agreement and disagreement of other philosophers in
thinking the issues through. The views that strike me as correct, with this
disagreement evidence set aside, are the views I should hold. Of course, the
evidence I get from agreement and disagreement remains epistemically rele-
vant to my all-things-considered beliefs. But for the purposes of the larger
project of which I am a part, in order to arrive at my views, I reason as if it is
not.12
A good way to get a handle on the proposal is to think about how one
typically reacts to a perceptual illusion, such as the one below.

Viewers almost always incorrectly judge the lettered regions to be different in

shade. Importantly, the apparent discrepancy between these identically
shaded regions tends to remain even after the viewer has become convinced

12While this paper is about philosophy, the idea may have broader application to other
collaborative, truth-seeking disciplines, such as in the Logic Team example.

-8-
of their constancy. The viewer continues to have the seeming or inclination,
but does not endorse it.13
An analogous phenomenon occurs when one gains evidence that one’s
own reasoning about a given topic is likely to be defective in some way. Con-
sider a case involving judgment-distorting drugs:14
Deducing While Intoxicated: Basil works through a non-trivial logic problem
and comes to believe p. She then learns that, before she attempted to solve the
problem, she ingested a drug that impinges on one’s deductive reasoning skills.
It causes ordinarily reliable thinkers to miss certain logic problems (such as the
one she just tried) at least half of the time. She rereads the problem and finds her-
self inclined to reason as before: The information given still seems to her to imply
p. But she refrains from endorsing this seeming and suspends belief.

In the story, Basil is, in some sense, inclined to accept a certain claim as true,
but opts not to endorse the inclination because of evidence that the mecha-
nisms that produced it may be epistemically defective in some way. This evi-
dence about one’s own cognitive capacities is widely known as ‘higher-order
evidence’ (evidence about one’s ability to evaluate evidence). Notice that the
‘seemings’ or ‘inclinations’ that persist despite what is learned are, in some
way, not sensitive to this higher-order evidence. In some sense, one can re-
tain the ability to see things as if the higher-order evidence were not there, or
were not relevant.
But how is this observation relevant to philosophy? Evidence from disa-
greement (and agreement) is thought to provide higher-order evidence, too.15
So the suggestion, to put it roughly, is this: Philosophers should favor the
views that seem right to them, ignoring certain bits of higher-order evidence
(including evidence from disagreement/agreement). David Chalmers helpful-
ly characterizes a related idea:16
[A] level-crossing principle... is a principle by which one’s higher-order beliefs
about one’s cognitive capacity are used to restrain one’s first-order beliefs about
a subject matter. [...] We can imagine a cognizer—call him Achilles—who is at
least sometimes insensitive to this sort of level-crossing principle. On occasion,

13 The seeming prompted by this illusion may involve alief, a representational mental state

that can conflict with one’s explicit beliefs. See Gendler (2008). For the purposes of this paper,
the operative attitude will be conditional belief, not alief.
14 See Christensen (2007, 2010, 2011, forthcoming) for thorough discussion of similar ex-

amples.
15 See Kelly (2005) and Christensen (2007) for influential early discussions that take this

viewpoint.
16 Others have made reference to an idea like this as well. Schoenfield (2014, pp. 2-3) de-

fines your ‘judgment’ as ‘the proposition you regard, or would regard as most likely to be cor-
rect on the basis of the first-order evidence alone.’ Horowitz and Sliwa (2015) make use of an
idea in this vicinity in their discussion of one’s ‘first order attitude’. While these attitudes are
close to the one I will rely on, the first-order/higher-order distinction turns out not to be quite
right for the purposes of this paper.

-9-
 Achilles goes into the mode of insulated cognition. When in this mode, Achilles
goes where first-order theoretical reasoning takes him, entirely insulated from
higher-order beliefs about his cognitive capacity. He might acquire evidence that
he is unreliable about mathematics, and thereby come to believe ‘I am unreliable
about arithmetic’, but he will go on drawing conclusions about arithmetic all the
same. (Chalmers 2012, pp. 103-104)

This idea of ‘insulated reasoning’ will be useful. The thought is that in phi-
losophy (and in other collaborative truth-seeking contexts), we should try to
reason in a way that is insulated from certain evidence, including the evi-
dence we get from disagreement, in determining our views. For reasons we
will discuss, it will turn out that we do not want to be insulated from all
higher-order evidence, as Achilles is in Chalmers’ example. But before we
can discuss which evidence we will want to wall ourselves off from, we will
need to say a bit about what ‘walling off’ or ‘insulation’ amounts to. As it
stands, it is unclear from Chalmers’ discussion whether insulated reasoning
is supposed to be something that we humans ever do or are even capable of.
The relevant sort of reasoning is of a quite familiar variety: conditional
reasoning. We reason conditionally when we reason as if, or supposing, or on
the condition that our evidence were different than it actually is.17 Conditional
reasoning can be divided into additive and subtractive varieties. In additive
cases, we introduce a supposition over and above whatever evidence we al-
ready have, and then reason accordingly (e.g. ‘Supposing we get to the DMV
by ten, I think it’s likely that we’ll be out of there by noon.’). Subtractive cases
are less common. In such cases, we focus only on a subset of our evidence,
ignoring or bracketing some of the evidence we already have, reasoning as if
it were not there.18 Trial jurors are sometimes expected to engage in this kind
of reasoning when a piece of evidence is deemed inadmissible. 19 Suppose, for
example, that unambiguous video footage of a defendant’s crime was collect-
ed without a search warrant. A juror who becomes aware of the video might
well be instructed to assess the likelihood that the defendant is guilty – set-
ting the video evidence aside altogether. Of course, the verdict she reaches
may differ markedly from her all-things-considered opinion about the guilt
of the defendant.

17 There is some impulse to understand claims of the form ‘Conditional on evidence E, I

think p’ counterfactually: ‘Were I to acquire evidence E, I would think p’ This impulse should
be resisted. An example from Horowitz (2015) can be used to illustrate this. Imagine that Ivan
becomes severely irrational whenever he believes that spiders are nearby. But let us add, real-
istically, that Ivan can suppose that there are spiders nearby without any significant problems.
What Ivan does think, on the supposition that there are spiders nearby, may turn out to be
quite different from what he would think, were he to learn that spiders are nearby.
18 It is worth noting that subtractive conditional reasoning is, formally, more problematic

than the additive conditional reasoning. We understand conditionalization; can we make sense
of ‘de-conditionalization’? See Joyce (1999) and Christensen (1999).
19 Thanks to an anonymous referee for suggesting this example.

-10-
 In both additive and subtractive cases, a person evaluates which way
some body of evidence (which may or may not be the evidence she in fact
possesses) seems to her to point. For ease of expression, if a certain body of
evidence E seems to me to support p, we will say that I am inclined, on E, to-
ward p.
We can now put the driving thought this way: When one is doing philos-
ophy with the aim of determining her philosophical views, she should not be
evaluating all of the evidence she has. Instead, she should be focusing on a
special subset of this evidence. Her views should be determined by her incli-
nations on this evidence (i.e. by the way this evidence seems to her to point).
What is the special subset of evidence? Provisionally, we will say that the rel-
evant subset is: All of the evidence minus that from disagreement and
agreement of fellow philosophers. So, on this picture, a person’s philosophi-
cal views should be her disagreement-insulated inclinations (i.e. the positions
that seem to her supported by the evidence not from agreement and disa-
greement of her peers). To see how this should work, we can apply it to the
Turning Tide example that caused some trouble for the speculation picture.
At the end of the Turning Tide example, the available arguments and ev-
idence seemed to Pat to support Physicalism, but the field was dominated by
Dualists. On the current proposal, in arriving at her philosophical views, Pat
is supposed to think about all of the evidence except for the evidence from
disagreement and agreement. This body of evidence includes the arguments
and objections with which she is familiar, and which, on balance, seem to her
to support Physicalism. In other words, Pat’s disagreement-insulated inclina-
tion is toward Physicalism. This makes her a Physicalist.
This is not to say that she should believe Physicalism to be true, or even
that she should be more confident of Physicalism than she is of its negation.
If she had to bet on one of the two positions, it would be wiser for her to bet
on Dualism, given its popularity among philosophers whom she has good
reason to respect. So her all-things-considered confidence in Physicalism may
be rather low, since it is sensitive to all her evidence, which includes the evi-
dence from disagreement. But the key point is that, as a member of the philo-
sophical community, one’s holding of a view and one’s all-things-considered
level of confidence in the truth of that view can come dramatically apart.

5 Answering the Dilemma

We began this paper seeking an attitude that would enable us to practice phi-
losophy in a sincere, rational way. Disagreement-insulated inclination is an
appealing candidate for the elusive attitude we sought. This section explores
how well the attitude fares with respect to the sincerity and rationality desid-
erata, and discusses the role that these inclinations should play more broadly.
Sincerity: Can disagreement-insulated inclination support sincerely held
philosophical views? The Physicalism example discussed at the end of the
previous section provides some reason to think so. For Pat does seem to be a

-11-
sincere Physicalist – both before and after Dualism becomes the dominant
view. It should be recognized, however, that there is one way in which Pat’s
commitment is less than fully sincere: She regards Dualism as more likely to
be true, all things considered. Nonetheless, it seems clear that Pat’s commit-
ment to Physicalism remains sincerely held, in an important sense. When we
characterized the sincerity desideratum initially, we wanted it to be the case
that the views we hold would trace back to our own consideration of the is-
sues. We wanted our views to seem right to us, in some important sense. And
Pat’s commitment to Physicalism meets these conditions with flying colors.
So, at least in this case, disagreement-insulated inclination seems well suited
to support the desired sort of sincerity.
Of course, there is a real question about why disagreement-insulated in-
clination is able to support sincerity. After all, disagreement-insulated incli-
nation is a special case of subtractive conditional reasoning (i.e. reasoning
that involves bracketing some of one’s evidence). And subtractive conditional
reasoning does not generally foster sincerity of the relevant sort: Think of the
juror who judges the defendant to be innocent only because she bracketed
overwhelming video evidence ruled inadmissible during her deliberation.
This juror is inclined, on the admissible evidence, toward the conclusion that
the defendant is innocent. But, presumably, this juror’s inclination would not
be sincere, in the relevant sense. Why should we expect that, as a rule, our
philosophical inclinations will be more sincere than the juror’s inclination in
this example?20
In thinking about this, it is important to note that the view on the table is
decidedly not that any old use of subtractive conditional reasoning will de-
liver sincere inclinations. Given a body of evidence, one can take any chunk
of it and set it aside in one’s reasoning. When one does this, the inclinations
one has will certainly not always ‘seem right’ in any important sense at all. It
will matter crucially which evidence one sets aside. But, then, what makes
disagreement evidence so special? Why does bracketing evidence from disa-
greement tend to result in seemings with which we sincerely identify?
As it happens, this property is not unique to disagreement evidence: It is
a feature of higher-order evidence more generally. Very often, bracketing
higher-order evidence will tend to facilitate sincere inclinations of the rele-
vant kind. Recall the example of Basil and the judgment-distorting drug.
There, Basil arrives at p by reasoning herself through the logic problem di-
rectly. In the end, she declines to trust the output of this reasoning, because of
information about the likely effects of the drug she took. All things consid-
ered, Basil might well regard p and not-p as equally likely to be correct. But
even if she regards both propositions as equally likely, she does not neces-
sarily harbor identical attitudes toward the two propositions. Importantly,

20 Thanks to an anonymous referee for posing this question.

-12-
Basil still has access to her direct thinking about this question, which led her
to p. As a result, if asked to defend p, Basil will have something sincere and
substantial to say – even if it turns out to be incorrect. If asked to defend not-
p, she may not have much to say at all – and if she were to try, it might well
feel to her as though she is trying to trick her audience. The key point is that a
certain kind of sincerity remains connected to Basil’s first-order judgment,
even when she admits that the first-order judgment is likely to be mistaken.
Higher-order evidence leads one to doubt one’s own direct thinking about
some core body of evidence. But there can often be a kind of seeming pro-
duced by that direct thinking – whether or not one trusts it. Even after one
acquires evidence that the direct thinking is not reliable, the direct thinking
does not tend to vanish; it will still be there, and it may even be correct. And
when the direct thinking is still there, we will be in a position to advocate for
the conclusion it produced in a way that bears important hallmarks of sincer-
ity: The position will seem right to us in a psychologically gripping way, and
we will have something substantial to say in support of it, which we can see
no flaw in. In short, because disagreement is a species of higher-order evi-
dence, it seems plausible that bracketing disagreement will tend to support
the kind of sincerity that No Rational Belief seemed to place in jeopardy.21
Rationality: Let us now turn to rationality. There are two important ques-
tions to address here. First, is the attitude of disagreement-insulated inclina-
tion rationally assessable? And, second, can we be rational in taking this atti-
tude toward our controversial philosophical views? There is good reason to
think that the answer to both questions is ‘yes.’ Evidence from disagreement
is, in general, what precludes us from rationally believing our controversial
philosophical views. With the disagreement evidence set aside, being in-
clined toward a view can indeed be rational – so long as the view one is in-
clined toward is in fact the view that the remaining evidence supports. A
modified version of the judgment-distorting drug case discussed earlier can
be used to illustrate the operative rational norm.22
Deducing While Intoxicated 2: Basil and Sage both attempt to solve a challeng-
ing logic problem with correct answer p. Before attempting the problem, each lo-
gician learned that she ingested a drug that impinges on one’s deductive reason-
ing skills. During past experimentation, Basil and Sage discovered that they tend
to answer challenging logic problems correctly about half the time, after having
ingested the drug.
On this occasion, Basil correctly deduces p, while Sage loses track of a nega-
tion symbol, and arrives at ~p. After obtaining these results, both immediately
temper their confidence in their respective answers considerably, because they

21 Here, one might be tempted to ask: Why only bracket disagreement evidence in one’s
philosophical reasoning? Why not just bracket all higher-order evidence? These questions lead
to some interesting complexities, which will be addressed in §6.
22 See Christensen (forthcoming) for discussion of closely related questions.

-13-
 suspect that their logical reasoning faculties were off-kilter. Indeed, despite hav-
ing reached different answers, both logicians end up with the same level of con-
fidence in p: .5.

Basil and Sage ended up at the same place. But intuitively, we want to say
that Basil’s reasoning was totally rational and that Sage’s was not (since
Sage’s reasoning was based importantly on a mistake).23 Thinking about this
case in terms of inclination can help us to account for this intuition. Even
though Basil and Sage have the same attitude toward p, all things considered,
notice that setting aside evidence about the drug, Basil and Sage are inclined to-
ward different positions. (Basil might say: ‘If this drug’s a dud, then definite-
ly p.’ Sage might say: ‘If this drug’s a dud, then definitely ~p.’) And notice
that because the non-drug evidence really does support p, only Basil’s incli-
nation is fully rational. So it does seem to make sense to say that insulated
inclinations can be assessed for rationality.
We should think of this situation as analogous to the situation that we
philosophers are often in. With respect to our all-things-considered attitude,
philosophical disagreement may compel us to be at the same place: agnosti-
cism. But setting the evidence from disagreement and agreement aside, we
will still have our inclinations, and these will differ from person to person.
Some of us are like Basil, rationally inclined toward the position the relevant
evidence supports. Others of us are like Sage, mistakenly inclined toward
some other position. So it will not turn out that all philosophers who take the
attitude of disagreement-insulated inclination are fully rational in holding
their views. But so long as one’s inclination is directed toward the position
that is in fact supported by the relevant evidence (i.e. the evidence not from
disagreement), one can be rational in holding that position as one’s view.
While it may be irrational to believe one’s controversial philosophical views, it
is not necessarily irrational to be inclined toward them, setting disagreement
aside.
The Role of Inclinations: At this point, we have seen that the prospect of
holding controversial philosophical views in a sincere, rational way can sur-
vive the threat posed by disagreement – provided that we philosophers take
the right kind of attitude toward our controversial views. But if we adhere to
this guideline – if the attitude underlying our philosophical commitments is
that of disagreement-insulated inclination – one might wonder about the sta-
tus of any actions based on those commitments.
A person’s philosophical commitments are rarely wholly inert; they may
lead a person to make assertions or to pursue courses of action. This raises
difficult, important questions. If we are supposed to harbor two distinct atti-
tudes toward certain philosophical propositions – our all-things-considered

23Kelly (2010, pp. 122-124) discusses a case with a similar structure (though no drugs are
involved).

-14-
credence and our dispute-insulated view – on which should we rely when we
teach our classes, when we advocate for public policies, when we vote, or
when we directly influence public policy in other ways?24
Though it may not be possible to answer these questions in a uniform or
categorical manner, it will be instructive to consider several different con-
texts, to illustrate how these questions can be answered, and the kinds of
considerations that are relevant, in thinking through these issues. To start,
consider a medical example:
Doctors: Holly is one of fifty doctors who are all concerned with a specific ill-
ness. These doctors have two distinct responsibilities related to this illness: They
treat patients who are suffering from the illness, and they also perform research
into how the illness should be treated.
There are two main drugs available for treatment and research: X and Y.
Though there is a large body of data available concerning the efficacy of these
drugs, the data are not fully conclusive. Forty-five of the doctors are inclined to
think that X is more effective, but the other five – including Holly – are inclined
to think that Y is more effective. For medical reasons, it is not possible to admin-
ister both X and Y to a single patient safely. And, in addition, it is not possible for
a doctor to research more than one of these drugs at a time.
Holly knows all of this and is trying to determine both how she should treat
the patients she sees and where she should focus her research efforts.

Given certain assumptions about the above case, it seems clear that all the
doctors – Holly included – should administer X to their patients. To see this,
suppose that Holly does not see herself as any more likely to make accurate
assessments of drug efficacy than the other doctors. Indeed, in the past, when
there has been an absence of consensus among the doctors about which of
two drugs is most efficacious, the larger group has tended to be right nine
times out of ten. Given this track record, and given that a great majority of
the doctors judged X to be more effective on this occasion, it would seem ir-
responsible for Holly (or any other doctor possessing the same evidence) to
do anything other than administer X to an ailing patient.
At the same time, it may well make good sense for Holly to investigate
the efficacy of Y in her research. As a member of the research community,
Holly should do whatever will aid the group in its efforts to determine con-
clusively which drug is most effective. Toward this end, it may not be opti-
mal to have all fifty doctors devoting their research efforts to the same drug –
even if that drug is currently the most promising one. Instead, it may be more
efficient to have a majority of the doctors researching the most promising
drug, with the rest researching alternatives that still have a decent chance of
turning out to be the best.25 If we add that the doctors, as a rule, tend to pro-

Thanks to the Editors of Mind for raising these questions.

24

See Ch. 8 of Kitcher (1993) for a thorough defense of the epistemic advantages bestowed
25

upon a scientific community by this kind of diversity. One intuitive takeaway of Kitcher’s dis-

-15-
duce their best research when they are permitted to investigate whichever
drug sincerely seems to them to be the best, then it will make sense for Holly
to research Y (even though she will be administering X to her patients).
With this example in mind, we can consider another structurally similar
example, in the realm of public policy. Suppose that Gavin is an expert on
some issue of public importance and is assessing the merits of a policy, which
may soon be enacted. When he thinks about the issue directly, he finds him-
self inclined toward the view that the policy is likely to have a quite positive
impact. But, like Holly, Gavin is in the minority. The vast majority of other
experts hold that the policy will, if enacted, have a quite negative impact. In
the past, when Gavin has been in the minority like this, he has tended to be
wrong much more often than right. As a result, he adopts two distinct atti-
tudes toward the policy: He harbors a sincere, disagreement-insulated view
that the policy would be beneficial, while his all-things-considered opinion is
that the policy is more likely to be detrimental. When it comes time to take
action, on which of these attitudes should Gavin rely? Naturally, the answer
to this question will depend on the specific act he is contemplating.
If Gavin is the governor and thereby has the ability to enact or veto this
policy entirely on his own, it is clear that he should veto the policy. The case
is relevantly similar to the one involving Holly’s treatment of her patients.
Gavin, like Holly, should pursue the course of action that is likeliest to bring
about a favorable outcome, given all the information available. It would seem
irresponsible to enact unilaterally a policy that the vast majority of experts
deem likely harmful.
In contrast, suppose that Gavin and the other experts are all members of a
committee tasked with assessing the likely effects of the policy in question.
Here, it is important for Gavin to advocate for his insulated viewpoint, de-
scribing the direct considerations that incline him to think that the policy is
likely to bring about a favorable outcome. If Gavin is wrong, then, with any
luck, the rest of the committee will be able to identify direct considerations
that outweigh or undermine the ones he raised. Of course, there is some
chance that his advocacy will sway the group to favor the policy. As long as
the other experts on the committee are generally reliable in making these
sorts of direct assessments, this change of opinion ought to be more likely to
occur when Gavin happens to be right. If Gavin were to stay quiet in such a
case, then the committee would be less likely to correct itself.
Voting contexts raise still different issues. First, consider a jury that votes.
One reason for jurors to rely on their disagreement-insulated attitudes in de-
termining how to vote is that doing so will tend to make the group more ac-

cussion is that communities that rapidly approach consensus can be at a disadvantage, since
inevitably they will occasionally converge around a mistaken opinion – which turns out to be
quite inefficient. It is better for a community to hedge its bets, delaying convergence until the
situation is relatively clear.

-16-
curate. As we saw in the Logic Team example, a group can have a higher
probability of reaching the right verdict when the group members reason and
vote independently than if they all defer to the most reliable individual per-
son in the group.26 And insulation from disagreement can serve to enforce
this kind of independence, thereby making the group more accurate.27
Let us turn to citizen voting. Suppose that Gavin is now contemplating
whether to vote for or against the policy mentioned above. This is a particu-
larly complex case.28 On the one hand, there is some intuitive pressure to say
that if Gavin really does regard the policy as likely to be detrimental, all
things considered, on the basis of expert testimony, then he should vote
against the policy. On the other hand, the considerations about group accura-
cy discussed in the context of jury voting, which count in favor of insulation,
also seem applicable in this case. So there is a tension here that cannot be re-
solved in the abstract, without reference to the particulars of a given case.29
The context of teaching also raises difficult and important issues. In one’s
efforts to portray a controversial philosophical issue honestly and fairly to
one’s students, is it better to rely on one’s insulated attitude or on one’s all-
things-considered attitude? It is worth noting that one need not rely on only
one of these attitudes in all aspects of one’s teaching. For example, a philoso-
pher should not present her controversial views in the same way that an in-
troductory science teacher would present an established scientific theory: She
should make clear to the students that there are well-qualified philosophers
on both sides. But a teacher’s own inclinations can play a role in her teaching:
She may tell her students how things look to her, when she considers the
matter directly. And she may encourage the students to work out their own
views by directly considering the relevant arguments. Similarly, these atti-
tudes may figure in complex ways in other decisions a teacher might make
(e.g. in deciding what to put on the syllabus, or in designing exams).
There are certainly going to be many other contexts in which it may not

26 Condorcet’s (1785) jury theorem shows this. The Logic Team is, in a way, an application

of the theorem.
27 Insulation can prevent herding – congregating around a particular opinion – which is a

natural result of properly accounting for disagreement evidence. See Lackey (2013) for a dis-
cussion showing that independence in belief-forming contexts is a tricky issue to resolve.
28 One layer of complication stems from the fact that the body of voters might not properly

be regarded as a ‘truth-seeking body’ in the way that juries can be. Instead, we might think of
voting as a fair way to reconcile our incompatible preferences. On this picture, Gavin is enti-
tled to vote for the policy he would prefer to see enacted, quite apart from its likely conse-
quences for society at large. To circumvent this worry, we can suppose that Gavin would pre-
fer to enact the policy if and only if it is likely to have a positive impact, on the whole. See, e.g.,
Estlund (2008) for discussion of these issues.
29 For example, one issue that arises here is that of expertise. In citizen voting, a person

may not think that her fellow voters are particularly reliable with respect to the issue being
voted on. The strength-in-numbers consideration discussed previously is applicable only if the
other voters are sufficiently reliable, with respect to the disputed issue.

-17-
always be clear which of one’s attitudes one should rely on. This should not
trouble us. In a real-world medical context, the treatment of patients and the
research into potential treatments may not be cleanly separable (as was sup-
posed in the Doctors example). Such a case would make it difficult to deter-
mine whether and to what extent any of the doctors should be acting on their
insulated views. Real-world cases are messy, and we should not assume that
there is bound to be a clean way to handle them. Nonetheless, we can often
identify some of the considerations that are relevant to making these kinds of
determinations.
Broadly speaking, we have observed two distinct benefits that can be as-
sociated with relying on one’s disagreement-insulated attitude (rather than
on one’s all-things-considered attitude): First, a distinctive kind of sincerity
can be enabled. Second, the efficiency of a deliberative, truth-seeking body
can be enhanced due to the way that insulation can foster a valuable sort of
cognitive diversity. Which attitude a person may choose to rely on in a given
context will depend on which of these benefits can be obtained and on the
extent to which she values them in that context. A philosopher may place a
high value on sincerity; a doctor seeing patients may not.
Insulated inclination is not a general substitute for belief. It is a separate
item in our epistemic toolkit, to be used when the task before us makes its
use appropriate. I suspect that there is no simple rule precisely describing the
conditions that warrant reliance on an insulated attitude. But it does seem
that, in the context of doing philosophy, employing a certain kind of insulat-
ed reasoning does make good sense.30

6. Inclination, insulated from what?

I will conclude by exploring whether disagreement evidence is the only kind
of higher-order evidence from which our philosophical reasoning should be
insulated – for there is a compelling case to be made that our philosophical
reasoning should be insulated from at least some other kinds of higher-order
evidence as well.
While disagreement in philosophy can be used to argue persuasively for
No Rational Belief, it is by no means the only route to this conclusion. Sup-

30 It might be thought that the foregoing picture runs into difficulties when accounting for

philosophical assertions – particularly if it is assumed that warranted assertion requires

knowledge or rational belief. There are two important observations to make about this worry.
First, it should be noted that this is not a special problem for the insulated inclination picture
described here; it arises as soon as No Rational Belief is assumed. Second, it is worth pointing
out that there may be some divergence between philosophical norms of assertion and ordinary
norms of assertion. See Goldberg (2013b) and DeRose (forthcoming) for discussions that ex-
plore this idea. If norms of assertion are contextually sensitive in this way, it should be possi-
ble to generate norms of philosophical assertion that are compatible with the picture outlined
in this paper. One specific such suggestion is that, in the context of philosophy, an assertion is
warranted just in case it is based on a rational inclination, insulated from the appropriate evi-
dence.

-18-
pose that Erika, a committed Conciliationist, is philosophizing in her office.
She discovers a philosophical question that, as far as she knows, has never
been investigated by philosophers before. Since there is no disagreement
(that she is aware of), her disagreement-insulated inclination and her all-
things-considered opinion will match. After some reflection on this new
question, she finds that p seems right to her. But then she pauses, thinking
about how much confidence she should have in p, given that it seems right.
And it is not at all clear that Erika can, all things considered, regard p as like-
lier than its competitors. Here is one route Erika might take to a more agnos-
tic position on p:31
Expected Disagreement: Although there is no disagreement about whether p yet,
Erika expects there soon to be some. She thinks that this question is exceedingly
likely to provoke disagreement from philosophers whom Erika respects (though
she may not be able to predict exactly who will be on which side). She decides
that she needn’t wait for anyone to say the words ‘I disagree.’ In anticipation of
the disagreement, she divides her confidence equally between p and p’s soon-to-
be competitors.

The above reasoning can be transformed into a more general argument for
agnosticism about difficult issues in philosophy. So there might well be disa-
greement-independent reason to refrain from having much confidence in
some of one’s philosophical views. And if there is, then for certain philosoph-
ical questions (such as those which are likely to engender substantial disa-
greement), disagreement-insulated inclination cannot fully deliver on its
promise: It cannot allow us to arrive at firm, rational opinions about how the-
se philosophical questions should be answered.
One way to react is to recognize that ‘evidence from expected disagree-
ment,’ like evidence from actual disagreement, constitutes higher-order evi-
dence of a certain sort. (The knowledge that fellow philosophers are likely to
disagree with one, about a particular question, is some evidence that one’s
own thinking about this question is likely to be somehow mistaken.) Perhaps
the next move is to demand more insulation – insulation from all higher-
order evidence. Let’s call an inclination based solely on first-order evidence
‘fully insulated.’32 So on this new proposal, our philosophical views should
be our fully insulated inclinations. This would allow Erika to have the view
that p, since p seemed correct to her before considering certain higher-order

31 There are other higher-order routes to agnosticism in philosophy that make no essential
reference to actual disagreement. Ballantyne (2014) pursues one such route, which appeals to
the merely could-have-been disagreement of “counterfactual philosophers” (people who likely
would have disagreed with your philosophical views, had they chosen to pursue philosophy).
Frances (2016) discusses a different route, which involves being aware of one’s past philosoph-
ical failings. Yet another route may flow from thoughts about the difficulty of philosophical
questions.
32 Recall that this was the kind of insulated reasoning that Chalmers (2012) and others

seemed to have in mind.

-19-
worries. Unfortunately, this revised proposal is too simple. The problem is
that, sometimes, higher-order evidence does not seem to be evidence that
philosophers should set aside in their thinking about the issues.
Consider an analogy. Imagine that a member of an admissions committee
learns that she is prone to implicit bias: Her (perhaps fully insulated) inclina-
tion is, often, to regard male applicants as being more deserving of admission
than relevantly similar female applicants. This information constitutes high-
er-order evidence, since it concerns the committee member’s ability to com-
petently evaluate first-order evidence. But it does not seem to be evidence
that she should set aside, in arriving at an independent judgment about a
given applicant’s merit. Intuitively, she should attempt to compensate for her
bias to some extent. And this will preclude her being fully insulated.
Let us move to an example from philosophy. One can imagine learning
that, in weighing a theory’s elegance against a theory’s resistance to potential
counterexamples, one tends to overvalue one of these virtues, relative to the
other. This information constitutes higher-order evidence, since it is evidence
about one’s capacity to evaluate competing philosophical theories. But once
someone did discover this about herself, I think it quite natural to think that
she would be justified in compensating accordingly in reasoning about these
theories. If she knows that she ordinarily tends to overvalue elegance (say),
then she might respond by settling on a less elegant theory slightly more of-
ten than she otherwise would have. This does not seem immediately prob-
lematic. If anything, it would be problematic to ignore this information. So it
seems clearly acceptable, at least sometimes, not to bracket higher-order evi-
dence in our philosophical thinking.
From what, then, should our philosophical reasoning be insulated? In try-
ing to answer this question, we should think again about why insulation can
be useful in the first place. Insulation has in part been motivated by sincerity.
But in the previous section, we discussed a second advantage: It fosters cog-
nitive diversity. Plausibly, philosophy progresses most efficiently when vari-
ous distinct positions are being investigated. We could, perhaps, imagine a
variant of the Logic Team case, in which each team is trying to find a proof of
a theorem as quickly as possible. Other things equal, the team might well be
better off if its members are not all trying the same type of proof. One might
think that something similar could be true for philosophy too: If there is
something valuable out there to be discovered (an argument, a counterexam-
ple, etc.), a group might be more on the whole likely to find it if the group’s
members pursue many different paths.
Suppose that the above is right: Cognitive diversity is indeed desirable in
philosophy. This insight can help us to determine which higher-order evi-
dence to set aside in our philosophical reasoning. Specifically, if some higher-
order evidence tends to undermine cognitive diversity (when properly ac-
counted for), then that counts in favor of setting it aside. And if the higher-

-20-
order evidence guarantees cognitive uniformity, then that counts strongly in
favor of setting it aside. Does higher-order evidence ever tend to breed cogni-
tive uniformity in this way? It can. Evidence from disagreement is the most
straightforward case. When two groups of philosophers disagree, Concilia-
tionism will recommend that both sides suspend judgment – a uniform out-
come. Evidence from expected disagreement functions in a similar manner.
Like evidence from actual disagreement, it tends to support suspension of
judgment. If all philosophers were to account for the expectation of disa-
greement in their reasoning (even if their reasoning were insulated from ac-
tual disagreement), a uniform agnosticism would result – at least with re-
spect to certain difficult issues. So it makes sense to ensure that our
philosophical reasoning is insulated from disagreement and expected disa-
greement (and other kinds of higher-order evidence that would guarantee
uniformity, if properly accounted for).
But what about evidence about one’s own bias toward elegant theories? It
is worth noting that that evidence about one’s elegance bias is unlike evi-
dence from disagreement in an important respect: Correctly compensating
for it in no way guarantees uniformity: A disagreement between two philos-
ophers could easily persist, even after one or both of them compensated for
an elegance bias afflicting them. At the same time, compensating for a perni-
cious elegance bias can undermine diversity in at least some cases: If I am the
only one afflicted by the bias, then (before compensating) I might be inclined
toward certain (particularly elegant) positions that others find less plausible.
After compensating, I might come to dismiss those positions, too, making the
overall distribution of views more uniform. If diversity were all that counts,
then it would seem clear that I should not compensate (i.e. my reasoning
should be insulated from evidence of this bias).
But diversity is not all that counts. Yes, my bias might lead me to pursue
views that no one else would. But since the bias is a pernicious one, it is likely
making me less individually accurate. And individual accuracy counts, too.
So our concern for diversity should be tempered by our concern for accuracy.
One might worry, though, that a concern for accuracy could be used to
motivate taking account of disagreement evidence, as well. After all, taking
proper account of disagreement evidence also makes one more accurate in
the long run. Setting it aside makes one less accurate. (This was especially
dramatic in the Logic Team example.) In the case of both bias evidence and
disagreement evidence, insulating facilitates diversity at the cost of individu-
al expected accuracy. Why insulate in one case but not the other?
To see the difference, we should notice that there are three values we are
simultaneously trying to promote: (1) individual expected accuracy, (2) cog-
nitive diversity, and (3) sincerity. In trying to determine which of our inclina-
tions should underlie our philosophical views, we should consider the extent
to which each of these values may be promoted or undermined by a given

-21-
policy, were everyone to adopt it.
Start with evidence from disagreement. Ensuring that philosophers rea-
son in a way that is insulated from disagreement evidence will hurt every-
one’s individual expected accuracy (as insulation always does). But it will
enable sincerity to a great extent (as we saw in Turning Tide), and it will fa-
cilitate cognitive diversity to a great extent (for if everyone were to take ac-
count of disagreement evidence, complete uniformity – suspension of judgment
by all parties – would result). So while there is a cost associated with insulat-
ing from disagreement evidence, there are substantial benefits (such as the
avoidance of complete uniformity).
The case for insulating from bias evidence is weaker. With respect to val-
ues (1) and (3), accuracy and sincerity, the case for insulating from bias evi-
dence may be as strong as the case for insulating from disagreement evi-
dence: After all, insulating from bias evidence hurts personal expected
accuracy (just as insulating from disagreement evidence does), and insulating
from bias evidence may enhance sincerity (since bias evidence does consti-
tute a kind of higher-order evidence). But while insulating from bias evi-
dence may enhance cognitive diversity to some extent (depending on the de-
tails of the case), it will not enhance diversity to the extent that insulating from
disagreement evidence does. Properly accounting for bias evidence would
not guarantee uniformity in the way that accounting for disagreement evi-
dence would. Even if every philosopher learned tomorrow that she harbored
a bias toward elegant theories and compensated accordingly, it is very doubt-
ful that consensus would result. So there are reasons to favor insulating from
disagreement over insulating from biases.
This is not to offer a clean-cut rule telling us precisely when to insulate
and when not to. But it is not obvious that any such rule should be given. I
am tempted to think that we philosophers might have some flexibility in cer-
tain cases, due to the fact that philosophers value sincerity to varying de-
grees. Suppose for the moment that both of the first two values (individual
expected accuracy and cognitive diversity) tell in favor of accounting for
one’s elegance bias (i.e. not reasoning insulated from it). Still, I think that it
might well make sense for a philosopher to adopt the view corresponding to
her inclination insulated from the bias. Perhaps this philosopher is better able
to defend her views when she finds herself sincerely committed to them. Or
perhaps she finds that she is most creative when she argues on behalf of
views that ‘seem right’ to her in the relevant sense. Even if, other things
equal, the community might prefer (on accuracy and diversity grounds) to
have a philosopher defending view A rather than view B, it could turn out to
be more valuable to the philosophical community to have a creative and per-
suasive advocate of B rather than a lackluster advocate of A. So while there
do seem to be some general observations we can make (e.g. insulating from
disagreement is typically a good idea), we may not be able to delineate the

-22-
bounds of insulation in philosophy with perfect precision. Indeed, there may
well be cases in which the philosopher has a choice: It will be acceptable to
insulate, and also acceptable not to. Since we are trying to promote several
different values that can diverge, this permissiveness is not a problem – in
fact, it is just what we should expect.

7 Conclusion
We began by assuming that philosophical belief in controversial claims was
irrational, in order to see what sense we could make of philosophy if this
were so. We saw that the ‘sincere’ philosopher – the philosopher who holds
views that seem correct to her – faces a dilemma: Either she may believe her
views irrationally or else abandon an important kind of sincerity underlying
her philosophically controversial commitments. This paper has proposed
that, if careful, the sincere philosopher can retain most of what she might
want. In thinking about philosophical questions, the sincere philosopher
should set some of her evidence aside – including the evidence provided by
disagreement. She can sincerely and rationally advocate the views that she is
inclined toward, with this evidence set aside. Though she may not believe her
controversial views, all things considered, she can hold views that seem cor-
rect to her, in an important sense.
This is the view I am putting forward for your consideration. I fully ex-
pect to encounter dissenting opinion, and I confess that this expectation pre-
vents me from having high confidence in the proposal, all things considered.
Nonetheless, I can say sincerely that it seems to me to be correct.33

References
Ballantyne, Nathan 2014: ‘Counterfactual Philosophers,’ Philosophy and Phe-
nomenological Research 88 (2): pp. 368–387.
Bourget, David and Chalmers, David 2014: ‘What Do Philosophers Believe?’
Philosophical Studies 170: pp. 465–500.
Brennan, Jason 2010: ‘Scepticism about Philosophy,’ Ratio 23: pp. 1–16.
Chalmers, David 2012: ‘Insulated Idealization and the Problem of Self-

33 Versions of this paper were at the Pacific and Eastern Division Meetings of the Ameri-

can Philosophical Association in April of 2016 and January of 2017. I am grateful to audiences
on both occasions for the helpful feedback they provided. For other valuable comments, I’d
also like to thank two anonymous referees at Mind, Austin Andrews, Nomy Arpaly, Anna
Brinkerhoff, Dan Campana, Keith DeRose, Jamie Dreier, Nina Emery, Dave Estlund, Sandy
Goldberg, Yongming Han, Josh Knobe, Nick Leonard, Han Li, Chad Marxen, Don Maier,
Jakob Reckhenrich, Josh Schechter, the third-year and dissertation workshops at Brown Uni-
versity, the students from Keith DeRose’s seminar on recent work in the philosophy of religion
at Yale University, which met in the spring of 2016, and the students from my epistemology
seminar at Wheaton College, which met in the fall of 2016. Special thanks to David Christen-
sen.

-23-
 Doubt,’ Constructing the World. Oxford University Press: pp. 101–107.
Christensen, David 1999: ‘Measuring Confirmation,’ Journal of Philosophy, 96:
pp. 437–461.
— 2007: ‘Epistemology of Disagreement: the Good News,’ Philosophical Re-
view 116: pp. 187–217.
— 2010: ‘Higher-Order Evidence,’ Philosophy and Phenomenological Research
81 (1): pp. 185–215.
— 2011: ‘Disagreement, Question-Begging and Epistemic Self-Criticism,’
Philosophers’ Imprint 11 (6): pp. 1–22
— 2013: ‘Epistemic Modesty Defended,’ in Christensen and Lackey, pp. 77–
97.
— 2014: ‘Disagreement and Public Controversy,’ in Lackey, J. (ed.) Essays in
Collective Epistemology. Oxford University Press: pp. 143–163.
— forthcoming: ‘Disagreement, Drugs, etc.: From Accuracy to Akrasia,’ Epis-
teme.
Christensen, David and Lackey, Jennifer (eds.) 2013: Disagreement: New Es-
says. Oxford University Press.
Condorcet, Marie Jean Antoine Nicolas de Caritat, Marquis de 1785: ‘Essay
on the Application of Mathematics to the Theory of Decision-Making,’ in
Baker, Keith Michael (transl. and ed.) Condorcet: Selected Writings. Bobbs-
Merril, 1976, pp. 33–70.
DeRose, Keith forthcoming: ‘Appendix C: Do I Even Knowo Any of This to Be
True?: Some Thoughts about Knowledge, Belief, and Assertion in Philo-
sophical Settings and Other Knowledge Deserts,’ The Appearance of Igno-
rance: Knowledge, Skepticism, and Context Vol. 2. Oxford University Press.
Elga, Adam 2007: ‘Reflection and Disagreement,’ Noûs 41 (3): 478–502.
Estlund, David 2008: Democratic Authority: A Philosophical Framework. Prince-
ton University Press.
Feldman, Richard and Warfield, Ted (eds.) 2010: Disagreement. Oxford Uni-
versity Press.
Frances, Bryan 2016: ‘Worrisome Skepticism about Philosophy,’ Episteme 13
(3): pp. 289–303.
Fumerton, Richard 2010: ‘You Can’t Trust a Philosopher,’ in Feldman and
Warfield: pp. 91–111.
Gendler, Tamar 2008: ‘Alief and Belief,’ Journal of Philosophy 105 (10): pp. 634–
663.
Goldberg, Sanford 2009: ‘Reliabilism in Philosophy,’ Philosophical Studies 124
(1): pp. 105–117.
— 2013a: ‘Disagreement, Defeat, and Assertion,’ in Christensen and Lackey:
pp. 167–189.

-24-
— 2013b: ‘Defending Philosophy in the Face of Systematic Disagreement’ in
Machuca, Diego E. (ed.) Disagreement and Skepticism. Routledge: pp. 277–
294.
Grundmann, Thomas 2013: ‘Doubts about Philosophy? The Alleged Chal-
lenge from Disagreement’ in Tim Henning & David Schweikard (eds.),
Knowledge, Virtue, and Action: Essays on Putting Epistemic Virtues to Work.
Routledge: pp. 72–98.
Horowitz, Sophie 2015: ‘Predictably Misleading Evidence,’ Conference on
Rational Belief and Higher Order Evidence, Brandeis University.
Joyce, James 1999: The Foundations of Causal Decision Theory. Cambridge Uni-
versity Press.
Kelly, Thomas 2005: ‘The Epistemic Significance of Disagreement’ in Haw-
thorne, J. & Gendler, T. (eds.), Oxford Studies in Epistemology, Volume 1.
Oxford University Press. pp. 167–196.
— 2010: ‘Peer Disagreement and Higher Order Evidence,’ in Feldman and
Warfield: pp. 111–174.
Kitcher, Phillip 1993: The Advancement of Science: Science Without Legend, Ob-
jectivity Without Illusions. Oxford University Press.
Kornblith, Hilary 2010: ‘Belief in the Face of Controversy,’ in Feldman and
Warfield: pp. 29–52.
— 2013: ‘Is Philosophical Knowledge Possible?’ in Machuca, Diego E. (ed.)
Disagreement and Skepticism. Routledge: pp. 260–276.
Lackey, Jennifer 2013: ‘Disagreement and Belief Dependence: Why Numbers
Matter’ in Christensen and Lackey: pp. 243–268.
Licon, Jimmy Alfonso 2012: ‘Sceptical Thoughts on Philosophical Expertise.’
Logos and Episteme 3 (3): pp. 449–458.
Schoenfield, Miriam 2014: ‘A Dilemma for Calibrationism.’ Philosophy and
Phenomenological Research 90 (2): pp. 1–31.
Sliwa, Paulina and Horowitz, Sophie 2015: ‘Respecting All the Evidence.’
Philosophical Studies 172 (11): pp. 2835–2858.

-25-